A method for generating at least one feature for an occupant protection system based on a detected acceleration signal usually receives acceleration data continuously after initialization of an associated control unit. By analyzing such acceleration data, this method determines whether the vehicle is in a normal driving situation or in a potential crash situation at a given moment. A crash situation is referred to as a potential crash situation because the method cannot always decide reliably up to this point in time whether it is an actual crash or a case of misuse. During the normal driving situation, large portions of the method are in a dormant state or in a reset state, i.e., no calculations are performed, thus allowing an approach that conserves computer capacity. When a potential crash incident is detected, the parts of the method that are in a dormant state are also activated and, for example, features such as vehicle deceleration and/or forward displacement of occupants during the crash are calculated, threshold value curves are analyzed, functions are calculated pertaining to triggering decisions, etc.
If, after leaving the dormant state, it is found that it is only a misuse situation, it is necessary to ensure that the calculated features and variables, such as the deceleration and forward displacement of occupants, are restored to their starting state after the end of the misuse situation.
Known methods for generating at least one feature for an occupant protection system frequently use window integrators that calculate the features of deceleration and occupant forward displacement based on a detected acceleration signal. The window integrators are automatically reset to a starting value when no significant directional acceleration signal is applied for a prolonged period of time, i.e., the window integrators have a memory limited in time by the length of the window. In an alternative approach, the integrators are started for calculation of the features of deceleration and occupant forward displacement only when the detected acceleration signal reaches or exceeds a so-called noise threshold. If this noise threshold is no longer exceeded by the detected acceleration signal for a prolonged period of time, the integrators are reset again and/or reintegrated over a constant value or reset immediately to a starting state.